Do We Know How to Tame “Warrior” Cops?

How hard is to reform police forces to become less violent and less confrontational? Connie Rice is a civil rights attorney who has been working on this problem for years in Los Angeles, and in the LA Tmes today she says there’s some good news:

Under warrior policing, communities on the right side of the “thin blue line” receive safety and protection; communities on the wrong side receive suppression and prison. The good news is, we know how to change this.

In 2010, then-LAPD Chief Beck and activists created the Community Safety Partnership, a holistic, problem-solving approach to safety for high-crime areas that minimizes suppression, maximizes trust and acts through partnerships among residents, gang interventionists, local leaders, experts and other agencies to remove the root causes of trauma and crime. A recent UCLA evaluation found that reductions in violent crime and gang control in these partnership sites exceeded countywide declines, and the gains were accomplished with far fewer arrests and no police shootings or beatings. Moreover, residents reported higher trust in Community Safety Partnership officers.

Residents of neighborhoods traumatized by gang violence, predatory policing, endemic poverty and mass incarceration do not report seeing significant improvement in policing since 1992. The exception was residents in partnership sites.

Here’s what makes this kind of thing so difficult to evaluate. What Rice says about the UCLA report is true, but it relies on an analysis of only two neighborhoods, which it then compares to a “synthetic” neighborhood created to be statistically similar to the neighborhoods under study. This methodology can be fine, but it’s also really easy to misuse: the accuracy of the synthetic neighborhood is critical to the results. And in this case, it turns out that almost all of the “improvement” in the real neighborhoods is a bit of a mirage. Nothing much happened for years, but in 2015 crime rates in the synthetic neighborhood suddenly went up, which made the real neighborhoods look better:

Now compare this to a report from the Urban Institute that was commissioned by the Housing Authority of Los Angeles, one of the partners in the CSP initiative. It looked at seven neighborhoods and compared them to seven similar neighborhoods rather than constructing a single synthetic control. Here’s their evaluation:

Overall, we found that CSP reduced crime and improved police-community relations in at least one housing development.

Oof. That doesn’t sound so promising. Let’s look a little further:

Crime rates were consistently higher in CSP developments than comparison developments, but trends are similar across CSP developments and comparison developments. No clear patterns before or after any CSP wave exist. Thus, it is difficult to determine from these figures whether CSP has impacted crime, underscoring the need for a more rigorous methodological approach.

But wait! There’s also this:

The CSP neighborhoods saw a 14 percent drop in crime compared to the control neighborhoods. That’s good. At the same time, if you look at the chart you can see that this result is driven almost entirely by the year 2018. That’s sort of a thin reed. And there’s this:

Interviews revealed that CSP officers generally perceived community trust in the police to have increased. However, HACLA managers offered mixed responses regarding residents’ perceptions of the police. Overall, residents generally do not trust the police and expressed concerns about mistreatment, including a lack of anonymity when reporting crimes. There were also concerns that CSP officers may have been acting contrary to the program’s goals.

In other words, cops thought the program was working, but the residents of the housing developments were a lot less sure. The rest of the report has more details on this.

In one sense, all I’m doing here is illustrating, as usual, how hard it is to do good social science research. It’s really complicated and slightly different methodologies can produce big changes in results. In this case, the UCLA report is optimistic about CSP, but it relies on a tiny sample size of two neighborhoods and shows only that its precise choice of a synthetic control suffered a synthetic increase in synthetic crime. The Urban Institute report uses a bigger sample set and real controls, but even so shows statistically significant results in only one category, mostly driven by the results of a single year. On the qualitative side, we see something similar: the UCLA report says the CSP program is viewed pretty positively, while the Urban Institute report is far more qualified.

So is it true that when it comes to the “warrior police” mentality, “we know how to change this”? I would say no. We have some intriguing hints in the CSP program, but not much more than that. This is a really hard problem.